Evaluation of border points

[HeyImJordi]

For raster points which are evaluated exactly at the boarder position, NaN values are included in evaluation so that the result is NaN, even the height at that raster point is desired as output

[CarawaySeed42]

Set to Milestone 1.3.0.
The issue is, that evaluating z, using crg_eval_uv2z, is always done using bilinear interpolation. If any of the z-values partaking in the interpolation is NaN then the resulting z is NaN, even if they have no influence on the result because their weight is zero.
Example:
0 x NaN = NaN
while
0 x NaN = 0
would be the desired behaviour but only if one if at least one of the partaking z values is not NaN and its weight is bigger than zero.